RF interference on external passthrough USB3 port #2

New issue

Open

opened 2026-09-25 19:39:47 +00:00 by dexterlb · 0 comments

dexterlb commented

2026-09-25 19:39:47 +00:00

Member

Capturing video from a USB3 capture card connected on the external USB port is unstable.

The issue is location-dependant: so far it has been reproduced in NTUST in Taipei and in the FOSDEM office at Kortenberg. We cannot reproduce the issue in Sofia.

Hypotheses:

The USB passthrough cable might be bad
The length of the USB passthrough cable might be equal to a harmonic of a specific RF signal that is present in these places
The power board's USB passthrough implementation might be bad
The ground loop caused by the power board's USB passthrough being grounded might be turning into an antenna
Certain USB sticks might be more susceptible to the issue than others

Experimental evidence:

At NTUST during COSCUP:
- Connecting the capture card directly to the radxa USB seems to fix/improve things? (@markvdb or someone who was there please verify)
At the FOSDEM office:
- Replacing the USB cable by a longer USB cable by another manufacturer seems to improve things by quite a lot
- Looping the USB cable through a E04SR241336A ferrite seems to improve things a bit

Things to try:

Remove the ground loop: use a second power board as just passthrough (only connect it to the usb cable and the capture card, and place it in a way that no part of it touches any metal part of the box)
Instead of the power board, use another USB3 passthrough thingy (neutrik?)
Instead of the power board, use a conventional USB3 extender (no more than 50cm long)
Power: make sure that this isn't a power issue (look at the usb's +5V rail with a scope - does it drop before the video stream crashes? maybe the capture card draws too much power? is there obvious noise on the power or ground rails?)
Swap the USB ports on the radxa - it would be really dumb if it is a port issue, but worth a try
Try a ferrite with better high-frequency rejection, like the 0461164181

Capturing video from a USB3 capture card connected on the external USB port is unstable. The issue is location-dependant: so far it has been reproduced in NTUST in Taipei and in the FOSDEM office at Kortenberg. We cannot reproduce the issue in Sofia. **Hypotheses:** - The USB passthrough cable might be bad - The length of the USB passthrough cable might be equal to a harmonic of a specific RF signal that is present in these places - The power board's USB passthrough implementation might be bad - The ground loop caused by the power board's USB passthrough being grounded might be turning into an antenna - Certain USB sticks might be more susceptible to the issue than others **Experimental evidence**: - At NTUST during COSCUP: - Connecting the capture card directly to the radxa USB seems to fix/improve things? (@markvdb or someone who was there please verify) - At the FOSDEM office: - Replacing the USB cable by a longer USB cable by another manufacturer seems to improve things by quite a lot - Looping the USB cable through a E04SR241336A ferrite seems to improve things a bit **Things to try**: - Remove the ground loop: use a second power board as just passthrough (only connect it to the usb cable and the capture card, and place it in a way that no part of it touches any metal part of the box) - Instead of the power board, use another USB3 passthrough thingy (neutrik?) - Instead of the power board, use a conventional USB3 extender (no more than 50cm long) - Power: make sure that this isn't a power issue (look at the usb's +5V rail with a scope - does it drop before the video stream crashes? maybe the capture card draws too much power? is there obvious noise on the power or ground rails?) - Swap the USB ports on the radxa - it would be really dumb if it is a port issue, but worth a try - Try a ferrite with better high-frequency rejection, like the [0461164181](https://www.tme.eu/be/en/details/0461164181/ferrites-for-cables/fair-rite/)

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

video/videobox#2

No description provided.

Rows
Columns